With the rapid development of multimedia technology and network technology, video information has widely apply to each domain. In the face of voluminous video information, how to find the content that oneself needs fast , effectively, have already become the urgent demand. Content-based video retrieval technique arises at the historic moment, becoming a new research field.This article on the basis of content-based video retrieval technique of the introduction, video process technology go on sure research and discussion search, to based on content video mainly. First of all,in the characteristic extracting in shot detection,we adopted HSV space that accords with the mankind perceived further,and improve the quantization method of HSV space.On this basis, adopt the weighting color histogram of divided nonuniformly as the difference characteristic among the frames. This method is on the basis of guaranteeing to search the requirement , reduce calculating etc., and can express the main content of the video better。Secondly, we analyzed the various methods of video shot detection,on the basis of twain–comparison which can detected out the abrupt transition and gradual transition at the same time , we improved its existing problem. in threshold chosen , adopted the self-adaptation threshold value based on partical statistics characteristic. The experiment shows : The method which this paper put forward can make better result. The ones that have improved the recall and precision. Finally , we do some discussion in key frame extraction and shot cluster. |